# Demographics and the Trilemma: How Population Aging Shapes Exchange Rate Regime Choice

## Abstract

This paper investigates how demographic structure influences the resolution of the monetary policy trilemma. Using panel data for 140+ countries from 1970 to 2020, we test whether population aging systematically reshapes trilemma configurations. We find that aging is most strongly associated with greater financial openness (Z1 = 3.42, p < 0.01) and marginally with higher exchange rate stability (Z1 = 0.63, p < 0.10), while the effect on monetary independence is null in the full panel and actually positive for OECD economies. Logit models confirm that aging modestly increases the probability of adopting a pegged exchange rate (Z1 = 0.79, p < 0.01), though this effect is substantially smaller in the expanded panel than in OECD-only estimates (Z1 = 3.14). The eurozone provides the strongest evidence: within-EMU demographic deviations powerfully predict current account imbalances (Z1_dev = -185.5, p < 0.001), with effects roughly 18 times larger than for OECD floaters. We extend this analysis to the CFA Franc Zone, Eastern Caribbean Currency Union, and Common Monetary Area, finding that the CFA zone -- young and demographically divergent -- shows amplification with the *opposite sign* (Z1 = +126.5, p < 0.001), confirming that removing the exchange rate absorber amplifies demographic imbalances generically, not just for aging European economies. A placebo test shows that non-union peggers exhibit zero amplification, confirming the mechanism is specific to monetary union membership. However, mediation analysis reveals that the trilemma channel does not meaningfully mediate the demographic current account effect -- demographics operate conditional on regime choice rather than through it. The findings imply that demographic heterogeneity within monetary unions creates strain that exchange rate flexibility would otherwise absorb, with projected imbalances reaching 9-10 percentage points of GDP for the most demographically extreme eurozone members by 2050-2060.


**JEL Classification:** F31, F33, F41, J11

**Keywords:** demographics, impossible trinity, exchange rate regime, eurozone, CFA franc zone, monetary union, financial openness, population aging

## 1. Introduction

The monetary policy trilemma -- the impossibility of simultaneously maintaining free capital flows, a fixed exchange rate, and independent monetary policy -- is a cornerstone of international macroeconomics (Mundell, 1963; Obstfeld and Taylor, 2004). Countries must choose at most two of these three objectives, and the choice has profound implications for macroeconomic adjustment, current account dynamics, and financial stability. Yet relatively little attention has been paid to how demographic structure shapes this choice.

This paper bridges two literatures: the macroeconomics of population aging and the trilemma framework. We hypothesize that aging populations may shift trilemma resolution through several channels. First, older populations hold more nominal assets and may prefer exchange rate stability to protect asset values (Kopecky and Taylor, 2022). Second, aging economies tend to be net creditors and may favor financial openness to deploy capital abroad. Third, aging may reduce the political constituency for monetary policy activism, as retirees benefit from low and stable inflation. However, as we show, these channels operate differently across income levels and institutional settings -- the simple prediction that aging favors pegging holds in the full panel but reverses within the OECD, where aging is instead associated with higher monetary independence.

We operationalize the trilemma using the Aizenman-Chinn-Ito (ACI) indices of monetary independence (MI), exchange rate stability (ERS), and financial openness (FO) (Aizenman, Chinn, and Ito, 2010). Demographics are captured using the three principal components of the age distribution (Z1, Z2, Z3) following the methodology of our companion papers.

Our key findings are:

1. **Aging primarily increases financial openness, with weaker effects on MI and ERS.** Z1 (the primary demographic factor, correlated with overall aging) strongly predicts higher financial openness (Z1 = 3.42, p < 0.01) and marginally predicts higher exchange rate stability (Z1 = 0.63, p < 0.10), but has no significant effect on monetary independence in the full panel (Z1 = 0.14, NS). The OECD subsample reveals a surprising sign reversal: aging is associated with *higher* MI (Z1 = 2.88, p < 0.01) and insignificant ERS effects (Z1 = -0.83, NS).

2. **Demographics modestly predict peg adoption.** Logit models confirm that aging increases the probability of choosing a pegged exchange rate (Z1 = 0.79, p < 0.01 in the full sample), though the effect is substantially smaller in the expanded 140-country panel than in OECD-only estimates (Z1 = 3.14). The Z-KAOPEN interactions are all significant (p < 0.01), indicating that financial openness amplifies the demographic peg effect.

3. **The eurozone powerfully amplifies demographic current account imbalances.** Among eurozone members, the demographic effect on current account balances is roughly 18 times larger than for OECD floaters (Z1 = -189.8, p < 0.001 vs. -10.6, NS), consistent with the loss of exchange rate adjustment under monetary union. Within-EMU demographic deviations explain 33% of CA variation.

4. **Mediation is null -- demographics moderate rather than mediate.** Controlling for MI and ERS does not meaningfully attenuate the demographic CA coefficient (only ~7% attenuation in the trilemma-restricted sample, where the base Z1 effect is itself insignificant). Demographics operate conditional on regime choice rather than through it: the Z1-eurozone interaction is significant in the OECD subsample (Z1 x eurozone = -3.70, p < 0.01).

5. **The amplification mechanism generalizes beyond the eurozone.** The CFA Franc Zone -- 14 young, non-OECD African countries pegged to the euro -- shows demographic amplification with the *opposite sign* (Z1 = +126.5, p < 0.001), driven by youth-related divergence rather than aging. A placebo test demonstrates that non-union peggers show zero demographic amplification (Z1 = -7.0, NS), confirming the mechanism is specific to monetary union, not exchange rate fixity per se. The ECCU and CMA show directionally consistent but statistically insignificant effects (underpowered). The CFA result is robust to additional controls (trade openness, KAOPEN, GDP per capita) but fragile to excluding CEMAC commodity exporters.

The paper contributes to the literature on the determinants of exchange rate regime choice (Levy-Yeyati and Sturzenegger, 2003; Shambaugh, 2004), the macroeconomic consequences of aging (Aksoy et al., 2019; Auclert et al., 2021), and the emerging work on demographic influences on international capital flows (Higgins, 1998; Fair and Dominguez, 1991).

## 2. Literature Review

### 2.1 The Trilemma

The trilemma framework originates with Mundell (1963) and Fleming (1962), formalized by Obstfeld and Taylor (2004). Aizenman, Chinn, and Ito (2008, 2010) constructed continuous indices measuring the degree to which countries achieve monetary independence, exchange rate stability, and financial openness, demonstrating that the trilemma constraint binds empirically: increases in one index are associated with decreases in one or both of the others.

The choice of exchange rate regime has been studied extensively. Levy-Yeyati and Sturzenegger (2003) develop de facto regime classifications. Shambaugh (2004) constructs a binary peg indicator. Klein and Shambaugh (2010) examine the consequences of regime choice for trade and capital flows. Yet the determinants of regime choice remain debated, with existing work focusing on optimal currency area criteria (size, openness, trade concentration) and political economy factors (democracy, central bank independence).

### 2.2 Demographics and Macroeconomics

The lifecycle hypothesis predicts that demographic structure affects aggregate saving, investment, and therefore current account balances (Modigliani and Brumberg, 1954; Higgins, 1998). Recent work has established robust links between demographics and interest rates (Aksoy et al., 2019; Kopecky and Taylor, 2022), fiscal sustainability (Auerbach and Kotlikoff, 1987), and capital flows (our companion papers in this series).

The connection between demographics and exchange rate regime choice is largely unexplored. The closest work is by Juselius and Takats (2021), who document that aging affects inflation dynamics, which could indirectly influence regime preferences. Our paper provides the first systematic empirical test of whether demographic structure predicts trilemma resolution.

### 2.3 Monetary Unions Beyond the Eurozone

The literature on non-European monetary unions provides important context for our out-of-sample tests. The CFA Franc Zone has operated since 1945, with the CFA franc pegged first to the French franc and then to the euro (Masson and Pattillo, 2005). Debrun, Masson, and Pattillo (2005) examine the welfare implications of West African monetary union, finding heterogeneous effects across members. The ECCU has maintained a fixed peg to the US dollar since 1976, while the CMA links Lesotho, Eswatini, and Namibia to the South African rand. These unions vary dramatically in their demographic profiles: the CFA zone is young and high-fertility, while the EMU is old and low-fertility. This variation provides a natural experiment for testing whether demographic amplification under monetary union is a generic phenomenon or specific to European aging.

### 2.4 Demographics and Capital Flows

Our companion papers establish that demographic principal components (Z1-Z3) robustly predict current account balances in both multilateral and bilateral settings. The present paper asks whether the exchange rate regime mediates this effect: do demographics influence current accounts partly through their effect on regime choice? As we show, the answer is no -- demographics operate conditional on regime choice rather than through it, with the eurozone providing the clearest evidence of regime-contingent demographic effects.

## 3. Data and Methodology

### 3.1 Data

We combine three main data sources:

- **Trilemma indices**: The Aizenman-Chinn-Ito (ACI) dataset provides continuous measures of monetary independence (MI, N = 7,263 obs across 154 countries), exchange rate stability (ERS, N = 9,836 across 160 countries), and financial openness (FO, N = 7,608 across 160 countries) from 1970 to 2020. Each index ranges from 0 to 1.
- **Exchange rate regimes**: The Ilzetzki-Reinhart-Rogoff (IRR) classification provides de facto regime categories (N = 9,851 obs), which we aggregate into a binary peg indicator (71.7% peg, 12.8% float, with the remainder intermediate).
- **Demographics**: Demographic principal components (Z1, Z2, Z3) and dependency ratios from UN World Population Prospects, covering 237 countries from 1950-2024, merged via our multilateral panel.
- **Controls**: Fiscal balance, NFA position, GDP growth, relative output per worker, and KAOPEN from standard sources (IMF WEO, Lane-Milesi-Ferretti, Penn World Table, Chinn-Ito).

Table 1 reports summary statistics. The full panel covers 237 countries over 1950-2024, with effective regression samples ranging from approximately 140-160 countries depending on variable availability. The trilemma sum averages 1.51 (SD = 0.45), confirming the binding constraint. The trilemma corner distribution is roughly even: 27.8% sacrifice MI (peg + open), 33.3% sacrifice ERS (float + open), and 38.9% sacrifice FO (closed + stable).

### 3.2 Empirical Specification

Our baseline specification is:

$$
Y_{it} = lpha_i + \gamma_t + eta_1 Z_{1,it} + eta_2 Z_{2,it} + eta_3 Z_{3,it} + \delta' X_{it} + arepsilon_{it}
$$

where $Y_{it}$ is a trilemma index (MI, ERS, or FO) for country $i$ in year $t$, $lpha_i$ and $\gamma_t$ are country and year fixed effects, $Z_{1-3}$ are demographic principal components, and $X_{it}$ is a vector of controls including fiscal balance, lagged NFA/GDP, real GDP growth, log relative output per worker, and KAOPEN.

We estimate this using Prais-Winsten panel GLS to account for serial correlation in the errors. For binary outcomes (peg choice, MI sacrifice), we estimate both linear probability models (LPM) and logit specifications.

For the mediation analysis, we follow a sequential approach:

$$
	ext{Step 1: } KAOPEN_{it} = lpha_i + \gamma_t + eta' Z_{it} + \delta' X_{it} + arepsilon_{it}
$$

$$
	ext{Step 2: } Y_{it}^{tri} = lpha_i + \gamma_t + eta' Z_{it} + 	heta \cdot KAOPEN_{it} + \delta' X_{it} + arepsilon_{it}
$$

If the Z coefficients attenuate when KAOPEN is included (Step 2 vs. a version without KAOPEN), then KAOPEN partially mediates the demographic effect on trilemma choice.

Similarly, for the CA mediation:

$$
	ext{CA}_{it} = lpha_i + \gamma_t + eta' Z_{it} + 	heta_1 MI_{it} + 	heta_2 ERS_{it} + \delta' X_{it} + arepsilon_{it}
$$

If the Z coefficients attenuate when trilemma indices are included, regime choice partially mediates the demographic CA effect.

### 3.3 Interaction Model

We also estimate a joint model with Z-trilemma interactions:

$$
	ext{CA}_{it} = lpha_i + \gamma_t + eta' Z_{it} + 	heta' T_{it} + \phi' (Z_{1,it} 	imes T_{it}) + \delta' X_{it} + arepsilon_{it}
$$

where $T_{it} = (MI_{it}, ERS_{it})'$. The interaction terms test whether the demographic effect on current accounts varies with exchange rate regime.

## 4. Results

### 4.1 Trilemma Index Results

Table 2 presents the baseline results for demographics predicting trilemma indices. The dominant finding is that Z1 strongly predicts higher financial openness (Z1 = 3.42, p < 0.01, N = 4,658, R-squared = 0.121), with economically large effects: a one-unit increase in Z1 is associated with a 3.4 percentage point increase in the FO index. The effect on exchange rate stability is positive but only marginally significant (Z1 = 0.63, p < 0.10, N = 4,482, R-squared = 0.061). The monetary independence result is null: Z1 = 0.14 (SE = 0.25, p > 0.5, N = 4,101, R-squared = 0.139), with no significant demographic effect on MI in the full panel.

This pattern -- aging strongly increasing financial openness but not significantly affecting MI or ERS -- suggests that demographics shape the trilemma primarily through capital account liberalization rather than through direct exchange rate regime preferences.

Table 3 reports the OECD/non-OECD split, which reveals striking heterogeneity. For OECD economies, Z1 is significantly associated with *higher* MI (Z1 = 2.88, p < 0.01) and insignificantly with *lower* ERS (Z1 = -0.83, NS) -- the opposite pattern from what the full panel suggests. Financial openness in OECD is actually negative (Z1 = -2.60, p < 0.10). Non-OECD economies drive the full-panel FO result (Z1 = 3.95, p < 0.01) and the ERS result (Z1 = 0.48, NS but directionally positive). The OECD sign reversal on MI implies that aging within advanced economies is associated with *retaining* monetary independence, possibly because aging creditor nations value the capacity for independent monetary policy to manage their asset portfolios.

Table 4 decomposes Z into old-age and youth dependency ratios. The old-age dependency ratio significantly reduces MI (old_dep = -0.66, p < 0.01) and ERS (old_dep = -0.81, p < 0.01), while increasing FO (old_dep = 0.63, p < 0.05). The youth dependency ratio significantly reduces ERS (youth_dep = -0.21, p < 0.01) and marginally reduces FO (youth_dep = -0.15, p < 0.10). Both dependency ratios lower ERS, which is directionally consistent with dependent populations being associated with less exchange rate stability.

### 4.2 Regime Choice

Table 5 presents the peg-vs-float logit analysis. In the full panel (N = 2,841, 119 countries), Z1 significantly predicts pegging (Z1 = 0.79, p < 0.01), with Z2 and Z3 also significant. The OECD-only estimate is substantially larger (Z1 = 3.14, p < 0.01, N = 673, 28 countries), suggesting that the demographic-peg relationship is amplified among advanced economies. The Z-KAOPEN interaction model shows that all three interactions are significant at the 1% level (Z1 x KAOPEN = 0.72, Z2 x KAOPEN = -0.10, Z3 x KAOPEN = 0.004), confirming that financial openness amplifies the demographic peg effect. These logit results are robust to region and income fixed effects (Table 12 in the phase 11 robustness analysis): Z1 remains significant at 1% across all specifications (from 0.15 with region + income FE to 0.79 pooled).

The ordered regime choice model tells a consistent story: Z1 significantly predicts lower probability of intermediate (Z1 = -0.65, p < 0.01) and floating (Z1 = -0.41, p < 0.01) regimes, confirming that aging pushes toward pegging.

The MI sacrifice variable is not significantly predicted by demographics in the full panel (Z1 = -0.46, NS, N = 4,101), consistent with the null MI result in Table 2. However, the age decomposition reveals that old-age dependency significantly predicts MI sacrifice (old_dep = 0.80, p < 0.01), and youth dependency does as well (youth_dep = 0.16, p < 0.05), suggesting opposing lifecycle effects that net to insignificance in the Z specification.

Table 12 compares logit and LPM specifications. For peg choice, the logit yields a larger Z1 coefficient (1.41, p < 0.01) than the LPM (0.86, NS), suggesting some nonlinearity. For MI sacrifice, both specifications yield null results for Z.

### 4.3 Eurozone Stress Test

Tables 6-7 examine the eurozone subsample. Among the 19 eurozone members (N = 447 post-accession observations), the demographic effect on current account balances is dramatically amplified: Z1 = -189.8 (p < 0.001), compared to -10.6 (NS) for OECD floaters. This roughly 18-fold amplification is consistent with the loss of exchange rate adjustment: without an exchange rate valve, demographic pressures manifest more strongly in external balances. The eurozone R-squared is 0.242 vs. 0.307 for OECD floaters.

Within the eurozone, demographic deviations from the EMU cross-sectional mean powerfully predict CA deviations: Z1_dev = -185.5 (p < 0.001, R-squared = 0.330 with controls). Distance from Germany's demographic structure is also highly significant (Z1_dist_DEU = -180.0, p < 0.001, R-squared = 0.353). The pre-crisis vs. post-crisis split shows that the effect strengthens after the sovereign debt crisis: pre-crisis Z1 = -78.3 (NS, N = 115) vs. post-crisis Z1 = -107.8 (p < 0.01, N = 297).

The EMU counterfactual analysis trains a peg logit on OECD non-eurozone members and predicts regime choice for EMU members. The result is striking: 0 of 18 members are predicted to peg on average, with Finland having the highest mean P(peg) at 0.515. This suggests most eurozone members' demographics are inconsistent with their chosen regime.

Yield spread analysis shows that demographics do not significantly predict spreads vs. the Bund -- Z1 = 9.0 (NS, p > 0.6) on the 10-country EMU subsample. NFA and growth dominate yield spread determination within the eurozone.

### 4.4 Mediation Analysis

Table 13 reports the KAOPEN mediation results. Z strongly predicts KAOPEN in the first stage (Z1 = 6.49, p < 0.01, N = 5,014, R-squared = 0.320). Including KAOPEN attenuates the Z1 coefficient on ERS from 1.02 (p < 0.01) to 0.63 (p < 0.10) -- a 38% attenuation -- suggesting KAOPEN partially mediates the demographic effect on exchange rate stability. The MI mediation is ambiguous: Z1 on MI moves from 0.007 (NS) without KAOPEN to 0.14 (NS) with KAOPEN, both insignificant.

Table 14 reports the CA mediation results, which constitute the paper's core mediation test. In the baseline CA regression (N = 4,881, 162 countries), Z1 = -31.1 (p < 0.01). Adding MI and ERS controls attenuates Z1 to -27.0 (p < 0.05, N = 4,024), a modest 13% attenuation. However, the phase 8 diagnostic analysis reveals this attenuation is primarily driven by sample composition, not genuine mediation. When restricting to the trilemma-available sample without trilemma controls (M2), Z1 drops from -31.1 to 6.4 (NS) -- the loss of observations, not the inclusion of trilemma indices, drives the apparent attenuation. In the trilemma-restricted sample, adding MI and ERS moves Z1 from 6.4 to 6.0 (NS) -- only 7% attenuation, and the base effect is already insignificant. The mediation channel is therefore null.

The subsample mediation analysis confirms this pattern. Within the eurozone, adding trilemma controls barely attenuates Z1 (from -172.4 to -171.0, less than 1% attenuation). For OECD floaters, Z1 drops from 20.6 (NS) to 5.6 (NS). For non-OECD, the base effect is negligible (-0.66, NS).

### 4.5 Joint Model and Regime-Contingent Effects

Table 15 presents the joint model with Z-trilemma interactions. In the Z1-only interaction model, neither the Z1 x MI (-0.50, NS) nor Z1 x ERS (0.07, NS) interactions are significant. The full joint model with all Z-trilemma interactions yields significant Z1 x ERS (-69.0, p < 0.05) and Z2 x ERS (10.1, p < 0.05) interactions, suggesting that the demographic CA effect is conditioned by exchange rate stability, but interpretation is complicated by multicollinearity (the Z1 main effect flips sign to +17.2, NS).

The regime-contingent analysis (phase 8) provides cleaner evidence. In the full panel, the Z1 x eurozone interaction is positive but insignificant (1.09, NS, N = 6,352). However, restricting to OECD countries reveals a highly significant interaction: Z1 x eurozone = -3.70 (p < 0.01, N = 1,434), indicating that among comparable advanced economies, eurozone membership amplifies the demographic CA effect by 3.7 percentage points per unit Z1. This confirms that demographics *moderate* the regime-CA relationship rather than *mediating* it -- the mechanism is conditional on regime choice, not channeled through it.

The creditor-debtor asymmetry within the eurozone further enriches this picture: debtors show a much stronger demographic effect (Z1 = -225.9, p < 0.001) than creditors (Z1 = -68.5, NS), and the Z1 x creditor interaction is marginally significant (-2.50, p < 0.10).

### 4.6 Robustness

Tables 8-12 report robustness checks:

- **Lagged demographics** (Table 8): Using 5-year lagged Z yields similar results. For MI, Z1_lag5 = 0.26 (NS) vs. contemporary Z1 = 0.14 (NS) -- both null. For ERS, Z1_lag5 = 0.48 (NS) vs. Z1 = 0.63 (p < 0.10) -- lagged is slightly weaker. For FO, Z1_lag5 = 2.70 (p < 0.01) vs. Z1 = 3.42 (p < 0.01) -- both strongly significant. The FO result is robust to timing.
- **First differences** (Table 9): All first-differenced specifications yield null results for demographics (all Z p > 0.10). This is consistent with demographics operating as a slow-moving level effect rather than a change effect. The R-squared values collapse to near zero (0.001 to 0.003).
- **OECD split** (Table 10): The OECD and non-OECD samples show strikingly different patterns. OECD: MI Z1 = 2.88 (p < 0.01), ERS Z1 = -0.83 (NS), FO Z1 = -2.60 (p < 0.10). Non-OECD: MI Z1 = 0.08 (NS), ERS Z1 = 0.48 (NS), FO Z1 = 3.95 (p < 0.01). The full-panel FO result is driven entirely by non-OECD economies.
- **Pre/post GFC** (Table 11): The ERS result is significant only pre-GFC (Z1 = 1.40, p < 0.01 pre-GFC vs. -0.07, NS post-GFC), suggesting a structural break. FO is significant in both periods but stronger post-GFC (Z1 = 3.77, p < 0.01 post-GFC vs. 2.18, p < 0.01 pre-GFC). MI remains null in both periods.
- **Financial centers** (Table 10): Excluding financial centers (LUX, IRL, HKG, SGP, CHE, NLD, BEL) slightly strengthens the ERS result (Z1 = 0.83, p < 0.05 vs. 0.63, p < 0.10 in the full sample) and the FO result (Z1 = 3.86, p < 0.01 vs. 3.42, p < 0.01). MI remains null.
- **Logit vs LPM** (Table 12): For peg choice, the logit Z1 = 1.41 (p < 0.01) vs. LPM Z1 = 0.86 (NS, N = 3,456), suggesting some nonlinearity that the LPM misses. For MI sacrifice, both are null.
- **EMU robustness** (phase 11): The within-EMU Z_dev coefficient is highly stable under leave-one-out analysis (range: -129.3 to -213.1, all p < 0.001), winsorization (baseline -185.5, winsorized -180.4), and hub exclusion (excl. hubs: -117.3, p < 0.01; post-2010 only: -103.6, p < 0.01; excl. hubs + post-2010: -37.9, NS).

### 4.7 Beyond the Eurozone: Other Monetary Unions

The eurozone amplification result raises a natural question: is this a eurozone-specific phenomenon, or does eliminating the exchange rate absorber amplify demographic imbalances in *any* monetary union? We test this by extending the analysis to three additional monetary unions: the CFA Franc Zone (14 countries in West and Central Africa, pegged to the euro via the CFA franc), the Eastern Caribbean Currency Union (ECCU, 6 countries with usable data, pegged to the USD), and the Common Monetary Area (CMA, 4 Southern African countries linked to the South African rand).

This out-of-sample test is particularly powerful because the CFA zone is demographically the *opposite* of the eurozone: young, high-fertility, with divergence driven by differential rates of demographic transition rather than differential aging. If the EMU result is about removing the exchange rate absorber generically -- not something specific to European aging -- the CFA zone should show amplification with the opposite sign.

#### Within-Union CA Regressions

Table 16 reports within-union CA regressions estimated separately for each monetary union. The results confirm the generic amplification hypothesis:

- **EMU** (19 countries, N = 447): Z1 = -189.8 (p < 0.001). Aging demographics drive surpluses relative to the union mean.
- **CFA** (14 countries, N = 528): Z1 = +126.5 (p < 0.001). Youth-driven demographics drive deficits. The sign reversal is exactly as predicted.
- **ECCU** (6 countries, N = 109): Z1 = -82.9 (NS). Directionally consistent with EMU but statistically insignificant due to small sample.
- **CMA** (4 countries, N = 146): Z1 = -89.8 (NS). Similarly underpowered.

The EMU and CFA results are the identification-carrying unions, as anticipated. Within-union Z deviations confirm the pattern: EMU Z1_dev = -185.5 (p < 0.001), CFA Z1_dev = +150.3 (p < 0.01).

#### CFA Deep Dive: WAEMU vs CEMAC

Within the CFA zone, the result is driven by the CEMAC sub-group (Central Africa): CEMAC Z1 = +405.3 (p < 0.05, N = 201, 6 countries), while WAEMU (West Africa) is individually insignificant (Z1 = 16.8, NS, N = 327, 8 countries). This concentration in CEMAC -- which includes commodity exporters Gabon, Equatorial Guinea, Republic of Congo, and Chad -- warrants careful interpretation. Robustness check 7a (Table 17) shows that excluding oil exporters renders the CFA result insignificant (Z1 = 30.5, NS), while the oil-only subsample shows very large effects (Z1 = 480, p < 0.05). The CFA amplification thus reflects an interaction between commodity-driven external balances and demographic structure under a fixed exchange rate, rather than a pure lifecycle savings channel.

However, the result partially survives excluding only the most extreme outlier (GNQ): CFA excl GNQ yields Z1 = 65.0 (p < 0.05, N = 490). And critically, the CFA Z1 coefficient survives all additional controls: adding trade openness (Z1 = 133, p < 0.001), KAOPEN (Z1 = 134, p < 0.001), log GDP per capita (Z1 = 141, p < 0.05), and all three simultaneously (Z1 = 77, p < 0.05). This suggests the amplification is not simply a spurious correlation with openness or income.

The age decomposition confirms the youth-driven channel: within the CFA zone, youth dependency is significant (youth_dep = -30.4, p < 0.001) while old-age dependency is null (old_dep = 1.1, NS), the mirror image of the EMU pattern.

#### Pooled Union Analysis

Table 18 reports pooled regressions with union dummies and Z × union interactions. The separate union dummies reveal heterogeneous level effects: EMU membership is associated with higher CA (+2.8 pp, p < 0.05), CFA with lower CA (-2.7 pp, p < 0.05), and ECCU with substantially lower CA (-12.0 pp, p < 0.001). The Z1 × individual union interactions are not individually significant in the pooled specification, reflecting the opposing signs across unions (EMU negative, CFA positive) that cancel in aggregation.

#### The Placebo Test

The most important robustness check is the placebo comparison (Table 19). If the amplification mechanism is about monetary union specifically -- the irrevocable elimination of the exchange rate valve -- then non-union countries that merely peg their exchange rate should show no amplification. They can, after all, abandon their peg if demographic pressures become unsustainable.

The results are decisive:

- **Non-union peggers** (89 countries, N = 1,801): Z1 = -7.0 (NS, p > 0.7). Zero demographic amplification.
- **Non-union floaters** (74 countries, N = 909): Z1 = +54.2 (p < 0.05). Significant but moderate.
- **All union members** (43 countries, N = 1,230): Z1 = 15.0 (NS in pooled form, but significant within each union separately).

The non-union pegger null is the key finding. Simply fixing the exchange rate does not amplify demographic effects. The amplification requires the irrevocable commitment of monetary union -- the feature that distinguishes EMU, CFA, ECCU, and CMA from ordinary pegs.

#### Time Period Stability

The CFA amplification is stable across time periods (Table 20): pre-2000 (Z1 = 177, p < 0.05, N = 120), 2000-2012 (Z1 = 142, p < 0.001, N = 224), post-2012 (Z1 = 94, p < 0.05, N = 184). The coefficient magnitude declines over time but remains significant throughout, suggesting a persistent structural relationship.

#### CFA Regime Strain Projections

Using the within-CFA Z deviation coefficients and UN WPP demographic projections, we compute predicted CA deviations from the CFA mean through 2060. The projected strain is heterogeneous: by 2040, Chad (CA_dev = -3.6 pp) and Mali (-2.5 pp) face the largest demographically-driven deficit pressure relative to the CFA mean, while Gabon (+2.5 pp) and Guinea-Bissau (+2.1 pp) face surplus pressure. Cross-sectional Z1 dispersion within the CFA zone rises from 0.20 (2000) to 0.41 (2040), suggesting increasing demographic heterogeneity will strain the union over the coming decades.

### 4.8 Regime Strain Projections

Using the within-EMU Z deviation coefficients and UN WPP demographic projections, we compute predicted CA/GDP deviations from the EMU mean for each member through 2060. The projected strain is substantial but heterogeneous:

- **Largest projected deficits by 2050**: Italy (-9.6 pp), Spain (-6.2 pp), Greece (-5.1 pp), Portugal (-5.7 pp), France (-4.0 pp), and Germany (-4.7 pp). These represent demographically-driven CA deficits relative to the EMU mean.
- **Largest projected surpluses by 2050**: Cyprus (+8.5 pp), Slovakia (+7.1 pp), Lithuania (+5.0 pp), Latvia (+4.4 pp), and Ireland (+3.3 pp).
- **Total spread widens**: From 8.9 pp in 2020 to 14.1 pp by 2040 and 18.1 pp by 2050, driven by the baby-boom cohort entering old age.
- **Non-monotonic paths**: Some countries (e.g., Finland, Germany) show reversal after 2040-2050 as the baby-boom effect fades, while Southern European members face persistent worsening through 2060.

The forward counterfactual peg probability analysis shows that by 2050, only 3 of 18 members are predicted to have demographics consistent with pegging (Cyprus, Lithuania, Malta), with major economies like Germany (P(peg) = 0.008), France (0.010), and Italy (0.006) having near-zero peg probabilities. This reinforces the tension: the eurozone constrains members in a regime their demographics increasingly disfavor.

## 5. Conclusion

This paper investigates how demographic structure influences the resolution of the monetary policy trilemma. The results are more nuanced than a simple "aging favors pegging" story. In the full 140-country panel, aging primarily increases financial openness (Z1 = 3.42, p < 0.01), with only marginal effects on exchange rate stability (Z1 = 0.63, p < 0.10) and null effects on monetary independence (Z1 = 0.14, NS). The OECD subsample reveals a sign reversal: aging is associated with *higher* monetary independence in advanced economies, suggesting that aging creditor nations may value independent monetary policy to manage their growing external asset positions.

The peg-vs-float logit confirms that aging modestly predicts peg adoption (Z1 = 0.79, p < 0.01), but this effect is substantially smaller in the expanded panel than in OECD-only estimates, indicating that the OECD composition drove earlier stronger results. The mediation analysis yields a null: the trilemma channel does not meaningfully mediate the demographic current account effect. Demographics operate conditional on regime choice rather than through it.

The paper's strongest result concerns monetary union amplification. Within EMU, demographic deviations from the union mean predict current account deviations with extraordinary power (Z1_dev = -185.5, p < 0.001, R-squared = 0.330). This effect is roughly 18 times larger than for OECD floaters, consistent with the loss of exchange rate adjustment. The extension to other monetary unions confirms this is a generic mechanism: the CFA Franc Zone -- young, non-OECD, and demographically divergent -- shows amplification with the opposite sign (Z1 = +126.5, p < 0.001), driven by youth dependency rather than old-age dependency. If EMU amplifies aging-driven imbalances and the CFA zone amplifies youth-driven imbalances, the common thread is the irrevocable elimination of the exchange rate absorber. The placebo test is decisive: non-union peggers show zero demographic amplification (Z1 = -7.0, NS), confirming that the mechanism requires the institutional commitment of monetary union, not merely exchange rate fixity.

We note two important caveats. First, the CFA result is concentrated in CEMAC commodity exporters; excluding oil exporters renders it insignificant, suggesting the amplification operates through commodity-demographic interactions rather than a pure lifecycle channel. Second, the ECCU and CMA are too small to provide statistical power, leaving the out-of-sample validation reliant on the CFA zone alone.

For policymakers, the results suggest that demographic heterogeneity within monetary unions creates imbalances that exchange rate flexibility would otherwise absorb. As the baby-boom cohort enters old age, the eurozone will face increasing strain from demographic divergence -- with Southern European and core economies aging rapidly while peripheral members age more slowly. The forward counterfactual shows that by 2050, nearly all major eurozone economies' demographics are inconsistent with pegging. Similarly, the CFA zone faces rising demographic dispersion through 2040, with demographically divergent members (Chad, Mali, CAR) facing persistent deficit pressure relative to the zone mean. The policy implication generalizes: any monetary union that cannot deploy fiscal transfers, labor mobility, or wage flexibility to absorb demographic heterogeneity will face strain proportional to its members' demographic divergence. Future research should decompose within-union adjustment channels -- bilateral capital flows, wage adjustment, intra-union migration, and fiscal transfers -- to determine which can substitute for the exchange rate channel that monetary union eliminates.

## References

Aizenman, J., Chinn, M.D. and Ito, H. (2008). Assessing the emerging global financial architecture: Measuring the trilemma's configurations over time. NBER Working Paper No. 14533.

Aizenman, J., Chinn, M.D. and Ito, H. (2010). The emerging global financial architecture: Tracing and evaluating new patterns of the trilemma configuration. *Journal of International Money and Finance*, 29(4), 615-641.

Aksoy, Y., Basso, H.S., Smith, R.P. and Grasl, T. (2019). Demographic structure and macroeconomic trends. *American Economic Journal: Macroeconomics*, 11(1), 193-222.

Auclert, A., Bardoczy, B., Rognlie, M. and Straub, L. (2021). Using the sequence-space Jacobian to solve and estimate heterogeneous-agent models. *Econometrica*, 89(5), 2375-2408.

Auerbach, A.J. and Kotlikoff, L.J. (1987). *Dynamic Fiscal Policy*. Cambridge University Press.

Fair, R.C. and Dominguez, K.M. (1991). Effects of the changing U.S. age distribution on macroeconomic equations. *American Economic Review*, 81(5), 1276-1294.

Debrun, X., Masson, P. and Pattillo, C. (2005). Monetary union in West Africa: Who might gain, who might lose, and why? *Canadian Journal of Economics*, 38(2), 454-481.

Fleming, J.M. (1962). Domestic financial policies under fixed and floating exchange rates. *IMF Staff Papers*, 9(3), 369-380.

Frankel, J.A. and Rose, A.K. (1998). The endogeneity of the optimum currency area criteria. *Economic Journal*, 108(449), 1009-1025.

Higgins, M. (1998). Demography, national savings, and international capital flows. *International Economic Review*, 39(2), 343-369.

Juselius, M. and Takats, E. (2021). Inflation and demography through time. *Journal of Economic Dynamics and Control*, 128, 104117.

Klein, M.W. and Shambaugh, J.C. (2010). *Exchange Rate Regimes in the Modern Era*. MIT Press.

Kopecky, J. and Taylor, A.M. (2022). The murder-suicide of the rentier: Population aging and the risk premium. NBER Working Paper No. 30275.

Levy-Yeyati, E. and Sturzenegger, F. (2003). To float or to fix: Evidence on the impact of exchange rate regimes on growth. *American Economic Review*, 93(4), 1173-1193.

Modigliani, F. and Brumberg, R. (1954). Utility analysis and the consumption function: An interpretation of cross-section data. In K.K. Kurihara (Ed.), *Post-Keynesian Economics*. Rutgers University Press.

Masson, P. and Pattillo, C. (2005). *The Monetary Geography of Africa*. Brookings Institution Press.

Mundell, R.A. (1963). Capital mobility and stabilization policy under fixed and flexible exchange rates. *Canadian Journal of Economics and Political Science*, 29(4), 475-485.

Obstfeld, M. and Taylor, A.M. (2004). *Global Capital Markets: Integration, Crisis, and Growth*. Cambridge University Press.

Shambaugh, J.C. (2004). The effect of fixed exchange rates on monetary policy. *Quarterly Journal of Economics*, 119(1), 301-352.


### Companion Papers in This Series

Peters, B. (2026). Population Aging and the Fiscal Sustainability Trap: Expenditure Asymmetry, Debt Dynamics, and the Limits of the Interest Rate Channel. Working Paper.

Peters, B. (2026). The Demographic Erosion of Fiscal Leverage: Twin Deficits in an Aging World. Working Paper.

Peters, B. (2026). Why Feldstein-Horioka Correlations Vary: Demographics and the Savings Retention Puzzle. Working Paper.

